What a pack plant auditor checks that an ISO 9001 binder does not show

An ISO 9001 certificate tells you the supplier runs a documented quality system. It says nothing about the weld schedule that was live on the machine the day your tabs were joined, nothing about whether the torque driver on your busbar was verified this shift or last quarter, and nothing about whether anyone tested the ESD bench before an operator sat down at it. The evidence that predicts whether your pack comes back as a field failure lives at three stations, weld, torque, and ESD, and a buyer should ask for the same records at all three whether the supplier holds IATF 16949 or has never heard of it. The certificate is the gate. The line is the question.

I have read a lot of binders. The good ones arrive in a black three-ring, tabs labeled, internal audit reports from the last two cycles, a corrective action log with real closure dates, management review minutes with names attached. Somebody spent a week on that. And not one page of it tells me what the resistance welder did at 2:40 on a Thursday afternoon when a new operator was running my nickel tabs on a weld current that had drifted because nobody had dressed the electrodes since Monday.

So the binder gets fifteen minutes in the conference room while the coffee is still hot, and then I ask to walk. That is the whole method. Everything below is what I look at once we are standing on the floor, and the order I look at it in.

The binder gets fifteen minutes

What I am checking in those fifteen minutes is not compliance. It is whether the documents look like they were written for an auditor or for the people on the line. The internal audit reports tell me the most: if every finding is a document control finding and nothing has ever been found at a production station, either the plant is perfect, which is unlikely, or the internal audit is a paperwork exercise. Same with corrective actions. If the last twelve closures are all "retrained operator," the plant is not fixing causes.

Then I close it, and I ask one question before we stand up: who can change a weld parameter on the machine that will run my parts, and what do they have to do first? I want to hear a name, a role, and a step. If the answer is vague, the walk is going to be long.

Weld: pull-test values, not checkmarks

My first question at the welder is not "is this calibrated." It is "show me the parameters that are on the machine right now, and show me the sheet they came from." What I want to see is that the live schedule matches a controlled parameter sheet with a part number and a revision on it, that the schedule is locked so an operator cannot nudge the current to make a stubborn weld stick, and that there is a change log showing who touched it and when. Electrode dressing is the quiet one. Ask what the interval is and ask to see the record of the last few dressings. A welder that has not had its electrodes dressed is a welder producing joints outside the schedule you validated.

The record that actually matters is the pull test. Destructive pull testing on a coupon or a sacrificed unit, at a defined frequency, with results written down as numbers in grams-force or newtons against a lower limit, plus a rule for what happens when one fails. I ask for the pull tests from the specific days my build ran. And I read the handwriting. If the same number appears fourteen days in a row in the same hand, that log is a transcription exercise, not a test. Weld monitors or energy monitors that log every weld are better evidence than a paper sheet, but only if somebody is reviewing the output and there is a sign-off showing it.

Torque: the tool, the number, the re-check

The torque station is where I watch hands, not paper. Is the driver a DC nutrunner with a transducer that records rundown, or a pneumatic pulse tool and a wrist? Both can be fine, but only one of them gives you a record per fastener.

Three documents, in this order. The calibration certificate for the driver, matched to the sticker on the tool and still in date. The torque audit record, which is a separate re-check with a torque analyzer on a sample of already-tightened joints, not the same thing as calibration. And the work instruction at the station, which has to match the controlled revision in the office. I have found a station running a superseded instruction more than once, and it is always because the new one was filed and never posted.

Then I ask about the high-current joints specifically. Busbar and terminal fasteners are the ones that loosen under thermal cycling, and if the program calls for a re-torque or a thread-locking step, I want to see it on the traveler with a second signature. A single torque pass with no verification on a joint that carries the pack's full current is a decision somebody made, and I want to know who.

ESD: the log that was filled in at 6:55 a.m.

The wrist strap tester at the entrance to the bench should have a daily log, and the tester itself needs calibration, which people forget. The ionizer needs decay-time verification on a schedule. Floor and bench surface resistance needs to be measured at the point of use and the values recorded, not measured once at installation and then trusted for five years. The fixture and the welder need a ground, including the secondary side on a resistance welder, which is a different concern than a wrist strap and gets missed constantly.

The tell is the timing of the entries. A strap log filled in at the start of each shift, with an occasional failing reading and a note about replacing the strap, is a real log. A log where every entry in a week is in the same pen and every result is a pass is a log that was filled out Friday afternoon. I also look at whether the strap is actually on the operator's wrist in July, and whether the totes and bags at the bench are dissipative or just blue.

The electrical checks and the reject bin

Before shipment, I want to see an insulation resistance or hipot check on the assembled pack with a recorded value and a pass-fail limit, not just a green light. And I want cell-level open-circuit voltage and internal resistance sorting if the cells come in loose, or pack-level OCV and IR before it goes in the box. Whether cell sorting is worth the cost depends on your supplier's cell source and your matching requirement, and for a matched cell lot from a tier-one cell maker it may be redundant. That is a judgment, not a rule.

The reject bin is more informative than the test itself. Where do failures go? Is there a disposition, a rework traveler, a re-test after rework, a scrap record with a serial number? A bin that only ever contains scrap with nothing written on it is where problems go to be forgotten, and reworked packs that ship without a re-test are the ones that come back.

Work instructions, and who actually knows the process

Training records are easy to produce and not very useful. I do something less comfortable. I pick the newest person on the line and ask them to explain why the second weld pulse exists, or why the spacer goes in that orientation. If they can't, the process lives in the heads of the two people who have been there ten years, and your continuity risk has a retirement date.

I also look at the station itself: is the instruction at the bench the same revision as the controlled copy, in a language the operator reads, with the current part number on it. A station with a laminated sheet from the previous program taped to the bench tells me the documentation system and the floor have stopped talking to each other.

When the BMS or enclosure is built somewhere else

If the BMS comes from a subcontractor and the enclosure from somebody else, a clean audit of the pack plant has covered maybe half your build. The scope has to follow the value. Ask for the sub-tier list, the incoming inspection on the received BMS, and a specific answer to whether anyone powers it up and communicates with it before it goes into the pack. Ask what the change notification obligation is and when the last firmware or component change was communicated to the pack plant. If the answer to that last one is "we would have to check," you have no visibility into a large share of what you bought.

None of this replaces a certificate, and I want to be honest about where the certificate does work. In low-volume industrial and robotics programs, IATF 16949 is usually neither required nor maintained, because the volume does not justify it and the automotive customers are not there. Treating its absence as a disqualifier screens out good suppliers. Treating ISO 9001 as a substitute screens in bad ones. Neither document is a substitute for standing at the welder with a pull-test log in your hand.

And no audit is proof. One day is a sample, and a good system can have a bad Tuesday. If you cannot get on site, ask for time-stamped photographs of the station, the log sheet, and the machine parameters, taken that day, and treat the answer as a sample too.

I keep a one-page record list on my phone and I photograph each station. Those photographs go in the file next to the certificate. Six months later, if the field returns start, the first question the team asks is what the process looked like the day we approved it, and I would rather answer with a picture of a pull-test sheet than with a page from a binder.